Model and Architecture Sample Clauses

Model and Architecture. The high level architecture of EviNets is illustrated in Figure 3.8. For a given question, we extract potentially relevant information, e.g., sentences from documents retrieved from text corpora using a search system. Next, we can use an entity linking system, such as TagMe [70], to identify entities mentioned in the extracted information, which become candidate answers. EviNets can further incorporate additional supporting evidence, e.g., textual description of candidate answer entities, and potentially useful KB triples, such as types [186]. Finally, question, answer candidates and supporting evidence are given as input to the EviNets neural network. t t Let us denote a question by q, and {qt ∈ R|V |}, as a one-hot encoding of its tokens from a fixed vocabulary V . ai is a candidate answer from the set A, and we will assume, that each answer is represented as a single entity. For each question, we have a fixed set E = Etext ∪ EKB of evi- dence statements e(i), i = 1..M , and their tokens e(i). A boolean function mention : A × E → {0, 1} provides the information about which answer candidates are mentioned in which evidences. Individual tokens qt, ai, e(i) are translated into the embedding space using a matrix WD× |V |, where D is the dimension of the embeddings, i.e., qemb,t = Wqt, aemb,i = Wat and